Marathi is the 13th most spoken language globally with around 90 million native speakers. Despite the growing use of this language, sentiment and emotion (SE) analysis research remains limited. This lack of extensive research poses a challenge to understanding the text’s emotional and attitudinal content. The objective of this paper is to develop a hybrid model for SE analysis of low-resource languages like Marathi. The proposed hybrid model uses a combination of the Horse Herd Optimization algorithm (HHA), the Bidirectional Recurrent Neural Network (Bi-RNN), and the Ortony, Clore, and Collins (OCC) model to analyze opinions in Marathi texts. Our proposed model integrates OCC, which is an affective cognitive computing framework and serves as a foundation for emotion classification. Additionally, we use a version of HHA for hyperparameter optimization, followed by training a Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Gated Recurrent Unit (Bi-GRU) model for final SE classification. Based on the results, the proposed model variants OCC+HHA+Bi-LSTM and OCC+HHA+Bi-GRU achieved remarkable performance scores in both SE classification tasks. For the sentiment analysis task, the model variants achieved F1 scores of 93.76% and 96.79% respectively for the testing set. For the emotion classification task, the model variants achieved F1 scores of 98.42% and 97.89% respectively. These variants outperformed other text-processing models in terms of classification metrics.